st: New version of -metareg- on SSC

[Roger Harbord <roger.harbord@bristol.ac.uk>]

A new version of the -metareg- package is available from SSC, thanks to Kit 
Baum. To describe and install the package, type -ssc describe metareg- from 
Web-aware Stata. It requires Nick Cox's -matvsort- command, so if you decide 
to install it and don't already have -matvsort- installed type -ssc install 
matvsort-.

-metareg- performs random-effects meta-regression (meta-analysis regression) 
on study-level summary (aggregate) data. My thanks to Stephen Sharp for 
giving his permission to revise his -metareg- program (version 1.06 published 
January 1998 in STB-42 as package sbe23) and to issue the new version under 
the same name. The major revisions are:

(1) an option to apply a Monte Carlo permutation test to obtain p-values, as 
proposed by Higgins & Thompson (2004).

(2) an improved algorithm for REML estimation of the between studies 
component of variance tau-squared. This the Stata maximum likelihood routines 
to maximize the REML log-likelihood directly.

In addition the new version is written as a Stata estimation command and so 
allows use of many of Stata's post-estimation commands. It also supports 
several extra -predict- options.

The new version of -metareg- continues to support the syntax of the previous 
version, although this is not documented in the help-file to avoid further 
complicating the syntax diagram and confusing new users. Previous do-files 
that called -metareg- should therefore continue to work, although the results 
may not be identical due to (2) above.

More detail will be given in a planned future Stata Journal article, but I 
would very much welcome comments and suggested improvements before I submit a 
final version to Stata Journal. (However I will be away from my email from 25 
September until 11 October so please be patient if emailing me queries before 
then.)

For information on permutation tests in meta-regression, including 
interpretation of the results, see:

Higgins JPT, Thompson SG (2004).  Controlling the risk of spurious findings 
from meta-regression.  Statistics in Medicine 23:1663-1682.
http://www3.interscience.wiley.com/cgi-bin/abstract/108565626/ABSTRACT

My thanks to Rebecca Boehm whose posts to Statalist originally inspired me to 
revisit -metareg- and to Aijing Shang for useful comments and suggestions.

Roger.
----------------------------------------------------
Roger Harbord     mailto:roger.harbord@bristol.ac.uk
Department of Social Medicine, University of Bristol

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/